| Message ID | 20190717105156.15721-1-l.stach@pengutronix.de (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 6d19d8a3cec74a9680947ecb6abdeda38583110e |
| Headers | show |
| Series | ASoC: fsl_sai: mark regmap as fast_io | expand |
On Wed, Jul 17, 2019 at 12:51:56PM +0200, Lucas Stach wrote: > The regmap is only ever used to access MMIO registers, so it's fair > to say that register access is fast. Are you sure there's no caches that try to do allocations at runtime? The main thing fast_io does is change the locking to spinlocks which means we shouldn't do any allocations under lock, such as when updating the cache.
Am Mittwoch, den 17.07.2019, 12:06 +0100 schrieb Mark Brown: > On Wed, Jul 17, 2019 at 12:51:56PM +0200, Lucas Stach wrote: > > > The regmap is only ever used to access MMIO registers, so it's fair > > to say that register access is fast. > > Are you sure there's no caches that try to do allocations at > runtime? The main thing fast_io does is change the locking to > spinlocks which means we shouldn't do any allocations under lock, > such as when updating the cache. This driver is using REGCACHE_FLAT, which does its sole allocation on init and none at reg access time AFAICS. So this should be okay. Regards, Lucas
diff --git a/sound/soc/fsl/fsl_sai.c b/sound/soc/fsl/fsl_sai.c index 8593269156bd..812c94fbb160 100644 --- a/sound/soc/fsl/fsl_sai.c +++ b/sound/soc/fsl/fsl_sai.c @@ -771,6 +771,7 @@ static const struct regmap_config fsl_sai_regmap_config = { .reg_bits = 32, .reg_stride = 4, .val_bits = 32, + .fast_io = true, .max_register = FSL_SAI_RMR, .reg_defaults = fsl_sai_reg_defaults,
The regmap is only ever used to access MMIO registers, so it's fair to say that register access is fast. Signed-off-by: Lucas Stach <l.stach@pengutronix.de> --- sound/soc/fsl/fsl_sai.c | 1 + 1 file changed, 1 insertion(+)